Facing the AI challenge in radiology: Lessons learned from a regional survey among Austrian radiologists in academic and non-academic settings on perceptions and expectations towards artificial intelligence.

Aim: This study aimed to evaluate perceptions and expectations towards artificial intelligence (AI) applications in diagnostic radiology among radiologists across academic, non-academic and private practice settings in the Federal State of Styria, Austria. It also sought to determine how participant's characteristics and AI-specific knowledge might influence these views.
Methods: An online quantitative survey comprising 20 multiple-choice questions in German language was distributed via email to radiologists in outpatient and hospital settings throughout Styria in 2024.
Results: Out of 149 radiologists contacted, 66 responded. Of these, 75.4% reported having basic knowledge of AI, 13.8% indicated good to very good knowledge and only 10.8% had minimal AI-specific knowledge. The majority (84.4%) expressed willingness to use certified AI software in diagnostics. About half of the respondents (50.8%) believed that AI would not fully replace radiologists in the next 10-15 years, although 46.0% anticipated partial replacement. Additionally, 87.7% did not foresee a decrease in professional income due to AI integration. 64.6% anticipated improvement in diagnostic tasks through AI, with this expectation being significantly linked to an academic career (χ ²  = 8.97, p   =  0.01). However, opinions varied on AI's potential to outperform radiologists in diagnostics in the near future. There was no statistically significant relationship between participant's AI-specific knowledge and perceptions and expectations towards AI.
Conclusion: The study reveals a generally positive attitude towards AI among radiologists, with uncertainties about its future performance compared to human radiologists. Although AI is anticipated to positively influence workload without reducing income, there may be a discrepancy between these expectations and actual outcomes.
Competing Interests: The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
(© The Author(s) 2024.)